4 ms·
> The only reliable countermeasures are outside the LLMs but they restrain agent autonomy. Do those countermeasures mean human-in-the-loop approving actions ma
by Esophagus4 1y ago
> The only reliable countermeasures are outside the LLMs but they restrain agent autonomy.
Do those countermeasures mean human-in-the-loop approving actions manually like users can do with Claude Code, for example?
- veganmosfet 1y agoYes, adding manual checkpoints between the LLM and the tools can help. But then users get UI fatigue and click 'allow always'.